I'm way behind on this list, so this thread is probably dead (or was...)
but FWIW, I'd contend that sending a SIGIO for "some more room is available"
is inefficient at best.  When someone posted the "description" of async I/O
I would have expected that SIGIO be sent for the transitions I/O not possible
to I/O possible only.  So, I'd say we have a bug.

--sjg

>> The real question is: do other OSs send the SIGIO only if the buffer
>> goes from "no room available" to "some room available", or does it
>> send it when the buffer goes from "some room available" to "some more room
>> available".

>It seems that FreeBSD, OpenBSD, Linux and Solaris have choosen the first
>option, and NetBSD the second one.